IPCC Models Versus Sea Surface Temperature Observations During The Recent Warming Period

OVERVIEW

This post compares satellite-based Sea Surface Temperature (SST) anomalies to the hindcasts and projections of the multi-model mean of CMIP3 models. CMIP3 is the archive the IPCC used as the source of their models for AR4. The period being discussed runs from November 1981 to November 2011. This covers most of the recent warming period that began in the mid-1970s.

There are two modes of natural climate variability discussed in this post: the El Niño-Southern Oscillation (ENSO) and the Atlantic Multidecadal Oscillation (AMO). For those new to ENSO, refer to An Introduction To ENSO, AMO, and PDO – Part 1. And for those new to the AMO, refer to An Introduction To ENSO, AMO, and PDO — Part 2.

This post also illustrates the multiyear aftereffects of the 1986/87/88 and 1997/98 El Niño events on the Sea Surface Temperature anomalies of the Atlantic, Indian, and West Pacific Oceans. Those oceans cover approximately 67% of the surface area of the global oceans. I have presented the processes that cause the multiyear aftereffects of those ENSO events in numerous posts over the past few years, so they will not be discussed in detail in this post. For those interested in learning about those processes, I discussed them and illustrated them with time-series graphs and with animated maps of sea surface temperature anomalies and other variables, most recently, in a two-part series: ENSO Indices Do Not Represent The Process Of ENSO Or Its Impact On Global Temperature and Supplement To “ENSO Indices Do Not Represent The Process Of ENSO Or Its Impact On Global Temperature”.

NOTE: The data in this post have been adjusted for the effects of volcanic aerosols.

INTRODUCTION

In the recent series of posts that compare the IPCC hindcasts for 20th Century surface temperatures to observed surface temperatures (see here, here, here, and here), the only time period when models consistently agreed with observations was the late warming period, from 1976 to 2000. But even that is misleading, because it gives the incorrect impression that anthropogenic forcings such as Carbon Dioxide were responsible for the rise in surface temperatures. Illustrating the error in that assumption is relatively easy when Sea Surface Temperature anomaly data is adjusted for the impacts of major volcanic eruptions and when the global data is divided into two subsets: the East Pacific (coordinates of 90S-90N, 180-80W) and the Rest-Of-The-World (90S-90N, 80W-180). Refer to the map in Figure 1 for an illustration of those areas. And Figure 2 is a comparison of the Sea Surface Temperature anomalies for those two subsets.

Figure 1

HHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHH

Figure 2

HHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHH

DATA

The Sea Surface Temperature anomaly data used in this post is Reynolds OI.v2. It combines bias-corrected satellite observations for more complete coverage and in situ observations from buoys and ships. The Reynolds OI.v2 Sea Surface Temperature data covers the period of November 1981 to November 2011, or 30 years. The Reynolds OI.v2 data is available through the NOAA NOMADS website here. There is another reason why the Reynolds OI.v2 data is used in this post: Smith and Reynolds (2004) Improved Extended Reconstruction of SST (1854-1997)stated about the Reynolds OI.v2 data:

“Although the NOAA OI analysis contains some noise due to its use of different data types and bias corrections for satellite data, it is dominated by satellite data and gives a good estimate of the truth.”

The truth is a good thing.

We’ll also be using the multi-model mean of the Sea Surface Temperature data that was produced by the climate models in the CMIP3 archive, where CMIP3 stands for Phase 3 of the Coupled Model Intercomparison Project. CMIP3 is the archive the IPCC used as the source of climate model data for its 4th Assessment Report. The CMIP3 Sea Surface Temperature data, identified as TOS, is available through the Royal Netherlands Meteorological Institute (KNMI) Climate Explorer website, specifically at their Monthly CMIP3+ scenario runswebpage. We have discussed in the recent posts that the multi-model mean represents the natural and anthropogenic forced component of the IPCC’s climate model outputs. And during the period we’ll be evaluating, it is the IPCC’s contention that anthropogenic forcings are the cause of the rise in surface temperatures.

The last discussion about the data is how the adjustments were made to account for the volcanic aerosols. The observational and model mean data are adjusted for the effects of volcanic aerosols, which would have major impacts on how the data was perceived during and for a few years after the explosive volcanic eruptions of El Chichon (1982) and Mount Pinatubo (1991). To determine the scaling factor for the volcanic aerosol proxy, I used a linear regression software tool (Analyse-it for Excel) with global Sea Surface Temperature anomalies as the dependent variable and GISS Stratospheric Aerosol Optical Thickness data (Source ) as the independent variable. The scaling factor determined was 1.431. This equals a global SST anomaly impact of approximately 0.2 deg C for the 1991 Mount Pinatubo eruption. To simplify and standardize the adjustments I’ve applied the same scaling factor to both the observed Sea Surface Temperature data and the model outputs. And I used the same adjustments for all subsets. As you will see, it slightly overcorrects in some instances and under-corrects a little in others. But since the adjustments are the same for the model outputs and instrument-based observations, they have no impact on the trend comparisons.

EAST PACIFIC SEA SURFACE TEMPERATURE COMPARISON

Figure 3 compares the Sea Surface Temperature anomalies of the East Pacific Ocean (90S-90N, 180-80W) to the scaled Sea Surface Temperature anomalies of the NINO3.4 region of the equatorial Pacific (5S-5N, 170W-120W). NINO3.4 Sea Surface Temperature anomalies are a commonly used index of the frequency and magnitude of El Niño and La Niña events, and I’ve scaled them (multiplied them by a factor of 0.22) because the variations in Sea Surface Temperature in that area of the equatorial Pacific are about 4.5 times greater than those of the East Pacific Ocean. As illustrated, the Sea Surface Temperature anomalies of the East Pacific mimic the NINO3.4 Sea Surface Temperature anomalies.

Figure 3

Figure 4 compares the observed Sea Surface Temperature anomalies of the East Pacific to the CMIP3 Multi-Model Mean for the same coordinates. The first thing that stands out is the difference in the year-to-year variability. The observed variations in Sea Surface Temperature due to the ENSO events are much greater than those of the Multi-Model Mean. Keep in mind when viewing the model-observations comparisons in this post that the model mean is the average of all of the ensemble members. And since the variations in the individual ensemble members are basically random, they will smooth out with the averaging. The average, therefore, represents the forced component (from natural and anthropogenic forcings) of the models. And it’s the forced component of the model data we’re interested in illustrating and comparing with the observations in this post, not the big wiggles associated with ENSO.

Figure 4

The difference in the linear trends between the Multi-Model Mean and the observations is also extremely significant. That is the focus of this post. The linear trend of the Multi-Model Mean is 0.114 deg C per decade for the East Pacific Ocean. This means, based on the linear trend of the Multi-Model Mean, that anthropogenic forcings should have raised the East Pacific Sea Surface Temperature anomalies, from pole to pole, by more than 0.34 deg C over the past 30 years. But the observed Sea Surface Temperature anomalies have actually declined. The East Pacific Ocean dataset represents about 33% of the surface area of the global oceans, and the Sea Surface Temperature anomalies there have not risen in response to the forcings of anthropogenic greenhouse gases.

THE REST-OF-THE-WORLD COMPARISON

The Sea Surface Temperature anomalies and Multi-Model Mean for the Rest-Of-The-World (Atlantic, Indian, and West Pacific Oceans) from pole to pole are shown in Figure 5. The linear trend of the multi-model mean shows that the models have overestimated the warming by about 23%.

Figure 5

But even that is misleading, because the observed Sea Surface Temperature anomalies only rose in response to significant El Niño-La Nina events, and during the 9- and 11-year periods between those ENSO events, the observed Sea Surface Temperatures are remarkably flat. This is illustrated first in Figure 6, using the period average Sea Surface Temperature anomalies between the significant El Niño events, and second, in Figure 7, by showing the linear trends of the instrument-based observations data between the 1986/87/88 and 1997/98 El Niño events and between the 1997/98 and 2009/10 El Niño events.

Figure 6

HHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHH

Figure 7

HHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHH

As you will note, I’ve isolated the significant El Niño events of 1982/83, 1986/87/88, 1997/98, and 2009/10. To accomplish this, I used the NOAA Oceanic Nino Index (ONI) to determine the official months of those El Niño events. There is a 6-month lag between NINO3.4 SST anomalies and the response of the Rest-Of-The-World SST anomalies during the evolution phase of the 1997/98 El Niño. So I lagged the ONI data by six months and deleted the Rest-Of-The-World SST data that corresponded to the 1982/83, 1986/87/88, 1998/98, and 2009/10 El Niño events. All other months of data remain.

Note: The El Niño event of 1982/83 was counteracted by the volcanic eruption of El Chichon, so its apparent role in the long-term warming is minimal.

And what do the climate models show should have taken place during the periods between those ENSO events?

For the period between the 1986/87/88 and the 1997/98 El Niño events, Figure 8, the model mean shows a positive linear trend of 0.044 deg C per decade, while the observed linear trend is negative, at -0.01 deg C per decade. The difference of 0.054 deg C per decade is significant.

Figure 8

The difference between the linear trends is even more significant between the El Niño events of 1997/98 and 2009/10, as shown in Figure 9. The linear trend of the observations is basically flat, while trend of the models is relatively high at 0.16 deg C per decade.

Figure 9

Keep in mind that the model mean, according to the IPCC, represents the anthropogenically forced component of the climate models during the period of 1981 to 2011. Unfortunately for the models, there is no evidence of anthropogenic forcing in the East Pacific Ocean Sea Surface Temperature data or in the Sea Surface Temperature data for the Rest Of The World.

Let’s subdivide the Rest-Of-The-World data even more. This will illustrate why the Sea Surface Temperature anomalies between the significant ENSO events are flat.

THE NORTH ATLANTIC AND THE SOUTH ATLANTIC-INDIAN-WEST PACIFIC SEA SURFACE TEMPERATURE ANOMALY DATA

Figure 10 is a map that shows how the data for the additional discussions were subdivided. Basically, this was done to isolate the North Atlantic from the additional ocean basins in the Rest-Of-The-World data. And the observed Sea Surface Temperature anomalies for those two subsets are shown in Figure 11. As illustrated, the linear trend of the North Atlantic Sea Surface Temperature anomalies is significantly higher than the linear trend of the South Atlantic-Indian-West Pacific subset. This higher trend in the North Atlantic data is caused by the additional mode of natural variability known as the Atlantic Multidecadal Oscillation. And as we will see, the forced component of the models (the model mean) does not account for the additional variability in the North Atlantic attributable to the Atlantic Multidecadal Oscillation.

Figure 10

HHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHH

Figure 11

HHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHH

Note: The North Atlantic Sea Surface Temperature anomalies for datasets like the Atlantic Multidecadal Oscillation data are normally depicted by the coordinates of 0-70N, 80W-0. Here they include 0-90N, 80W-40E to capture the Mediterranean Sea and corresponding portion of the Arctic Ocean leftover from the other subsets. The additional surface area has little impacton the North Atlantic Sea Surface Temperature anomaly data presented here. But to differentiate it from the other versions of the North Atlantic data, I’ve called it “North Atlantic Plus” in the graphs.

“NORTH ATLANTIC PLUS” COMPARISON

The North Atlantic is the only ocean basin where the models underestimate the long-term trend of the satellite-era Sea Surface Temperature data. See Figure 12. (Also refer to Part 1 and Part 2of an earlier two-part post comparing the Reynolds OI.v2 Sea Surface Temperature dataset to the same CMIP3 Multi-Model Mean, but note that the data in those posts have not been adjusted for volcanic aerosols.) Based on the linear trends, the models have underestimated the warming of the North Atlantic by nearly 35%. Again, the North Atlantic has an additional mode of natural variability called the Atlantic Multidecadal Oscillation or AMO. It seems very obvious that the multi-model mean fails to hindcast and project this additional variability.

Figure 12

And for those interested, I’ve also provided graphs that compare the model mean and observed trends between the significant El Niño events. As shown in Figure 13, the models underestimate the warming that took place between the El Niño events of 1986/87/88 and 1997/98. And as illustrated in Figure 14, the models overestimated the rise in North Atlantic Sea Surface Temperatures between the 1997/98 and 2009/10 El Niño events.

Figure 13

HHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHH

Figure 14

HHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHH

Let’s take a look at the South Pacific, Indian, and West Pacific comparison. As many of you are aware, I like to save the best for last.

SOUTH ATLANTIC-INDIAN-WEST PACIFIC COMPARISON

Figure 15 compares long-term observed Sea Surface Temperature anomalies and the Multi-Model Mean for the South Atlantic, Indian, and West Pacific Oceans. This is basically the portion of the “Rest-Of-The-World” dataset that is not included in the “North Atlantic Plus” data. As illustrated, the trend of the Multi-Model Mean is about 62% higher than the trend of the observed data. That is, the forced component of the models has over predicted the rise in Sea Surface Temperature anomalies for this subset by a substantial amount.

Figure 15

But the long-term trends are again misleading. The South Atlantic-Indian-West Pacific Sea Surface Temperature anomalies only rise during the significant El Niño events of 1986/87, 1997/98, and 2009/10. Between those events, the Sea Surface Temperature anomalies drop.

Figure 16 compares the observed South Atlantic-Indian-West Pacific Sea Surface Temperature anomalies to the Multi-Model Mean between the 1986/87/88 and 1997/98 El Niño events. The anthropogenic forcings have driven the model-mean upwards during this period, but the linear trend of the observations show that Sea Surface Temperatures declined. And the difference of 0.093 deg C per decade is a major difference. But that’s small compared to the difference between the linear trends of the observations and the model mean for the period between the El Niño events of 1997/98 and 2009/10. That difference is almost 0.18 deg C per decade.

Figure 16

HHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHH

Figure 17

HHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHH

CLOSING COMMENT

As illustrated in the two earlier posts that use these same datasets (see here and here), the Multi-Model Mean of the CMIP3 coupled ocean-atmosphere climate models do not hindcast and project the Sea Surface Temperature anomalies in any ocean basin, when the data is presented on times-series basis and on a zonal mean (latitude-based) basis. (The model mean of the West Pacific subset may look good on a time-series basis, but not on a zonal mean basis.)

This post confirms the Multi-Model Mean (the forced component of the climate models) does a poor job of hindcasting and projecting the actual rise in global Sea Surface Temperature anomalies, when the data is broken down into two logical subsets: the East Pacific Ocean and the Rest-Of-The-World. The post also illustrates the very basic reasons for that rise.

The models used by the IPCC for their hindcasts and projections assume that anthropogenic greenhouse gases drove the rise in Sea Surface Temperature anomalies from November 1981 to present. This is illustrated by the model mean, which represents the forced component of the models.

But the Sea Surface Temperature anomalies of the East Pacific Ocean (90S-90N, 180-80W) have not risen in 30 years. Refer to Figure 18.

Figure 18

And for the Rest-Of-The-World (90S-90N, 80W-180), Figure 19, the Sea Surface Temperature anomalies only rose during, and in response to, the 1986/87/88, 1997/98, and 2009/10 El Niño events.

Figure 19

There is no evidence that anthropogenic greenhouse gases have had any impact on the East Pacific Sea Surface Temperature anomalies (90S-90N, 180-80W) or on the Sea Surface Temperature anomalies for the Rest Of The World (90S-90N, 80W-180).

ABOUT: Bob Tisdale – Climate Observations

SOURCES

The model mean data is found at the KNMI Climate Explorer Monthly CMIP3+ scenario runs webpage. The Reynolds OI.v2 Sea Surface Temperature anomaly data is available through the NOAA NOMADS website here. And the GISS aerosol optical depth data used to make the adjustments for volcanic aerosols can be found at the Stratospheric Aerosol Optical Thickness webpage, specifically this data.

About Bob Tisdale

Research interest: the long-term aftereffects of El Niño and La Nina events on global sea surface temperature and ocean heat content. Author of the ebook Who Turned on the Heat? and regular contributor at WattsUpWithThat.
This entry was posted in Model-Data Comparison SST. Bookmark the permalink.

7 Responses to IPCC Models Versus Sea Surface Temperature Observations During The Recent Warming Period

  1. Jim says:

    Bob,

    Plan to do another ‘C3′ post using your chart(s). I’d like to ask you a question via e-mail prior to starting the work on that post. Could you send me an e-mail so I can then reply with the question? Thx.

    Jim, C3 Editor – c3headlines@gmail.com

  2. Bob Tisdale says:

    Jim, thanks for the interest. Feel free to ask your question via a comment and if you’d like me not to publish the question, just mark it as such.

    Regards

  3. maurymw says:

    It will be true that All about load pull technique, its importance, application and uses for measurement of the power factor.

  4. Pingback: ON THE IPCC’s UNDUE CONFIDENCE IN COUPLED OCEAN-ATMOSPHERE CLIMATE MODELS – A SUMMARY OF RECENT POSTS | Bob Tisdale – Climate Observations

  5. Pingback: On The IPCC’s Undue Confidence In Coupled Ocean-Atmosphere Climate Models – A Summary Of Recent Posts | Watts Up With That?

  6. Pingback: Preview of CMIP5/IPCC AR5 Global Surface Temperature Simulations and the HadCRUT4 Dataset | Bob Tisdale – Climate Observations

  7. Pingback: Preview of CMIP5/IPCC AR5 Global Surface Temperature Simulations and the HadCRUT4 Dataset | Watts Up With That?

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s